Naturalistic adoption and deliberate practice use of an AI-based OSCE platform during national licensure preparation

AI-based tools for objective structured clinical examination (OSCE) preparation show promise under controlled evaluation, but whether learners voluntarily adopt them intensively during authentic high-stakes licensure preparation remains unclear. We deployed Clinical Performance eXamination with Medical Students’ Assistant for Training and Evaluation (CPX-MATE), a voice-based AI platform combining a real-time evaluator (RTE) that scores peer role-play encounters and a virtual standardized patient (VSP) for solo practice. Among 72 eligible senior medical students, 43 senior medical students used it voluntarily 10–60 days before the Korean National OSCE. Adoption was 97.7% (42/43), with 97.9% session completion. Students completed 3,042 sessions (2,807 RTE evaluations, 235 VSP encounters) over 745.1 h—a median of 16.0 h and 66.5 sessions per user. Across all within-student, same-chief-complaint RTE repeat-practice pairs ( n = 776), second-session automated scores were modestly higher than first-session scores (+1.67 points [95% CI 0.99–2.23]; P < 0.001), mainly in history taking. Interviews indicated CPX-MATE supported deliberate practice by reducing peer-feedback burden. Scalable AI-assisted OSCE preparation achieved naturalistic adoption under high-stakes conditions.